nmr in drug discovery. from screening to structure-based design of antitumoral...

Tesi Doctoral

NMR IN DRUG DISCOVERY. FROM SCREENING TO STRUCTURE-BASED DESIGN OFANTITUMORAL AGENTS

Ricard A. Rodríguez Mias

Departament de Química Orgànica

Facultat de Química

Universitat de Barcelona

Barcelona, Juliol 2006

1 NMR METHODS IN DRUG DISCOVERY:A HANDS ON EXPERIENCE

NMR in Drug Discovery 39

Work presented within this first chapter was carried out in Dr Pellecchia’s Lab. at the Burnham Institute -SanDiego (CA)- during an internship in 2002. Dr Pellecchia’s laboratory focus falls in the use and developmentof NMR methodologies in Drug Discovery; particularly on the design of low MW inhibitors of protein-proteininteractions. At the time of my visit the most relevant targets in the laboratory were a series of proteinsinvolved in the regulation, or rather miss-regulation, of apoptosis in tumoral cells. Inhibition of these proteins

is both an interesting and challenging strategy for cancer treatment, and is currently been pursued by anumber of research groups. From a practical perspective, such targets will also fulfill an educational goal asthey will allow us to get first hand experience on some of the techniques outlined in the introduction.

40 NMR in Drug Discovery

1.1 SELECTIVE LABELING SCHEMES: BIR3-XIAP DOMAIN A CASE STUDY

1.1.1 BIR3-XIAP DOMAIN AND ITS BIOLOGICAL RELEVANCE

Apoptosis, or otherwise programmed cell death is a delicately controlled phenomenon in pluricellularorganisms, with utter importance to development and tissue homeostasis. A variety of pro and anti-apoptoticproteins control this process through a fine network of signaling events; and very often, such opposing

signals coexist in a dynamic equilibrium.[1, 2] This balance is altered through stimuli such as cell damage orother types of stress, and ultimately determines cellular fate. Missregulation however, has been pointed outas the cause leading to several conditions, both when balance is tilted towards excessive cell death(degenerative diseases), or disproportionate proliferation (cancer).

Figure 3 Multiple pathways leading to caspase activation.( see text for more details) Extracted fromKaufmann et al.[2]

Caspases are an important family of distinct cysteine proteases that share the ability to cleave substratesafter aspartate residues (hence its name Cys-ASP-ases). They have a central role in the apoptotic processand are largely responsible for its signal transduction, amplification, and execution. In fact one could regard

the entire network of protein-protein interactions controlling cellular suicide as a sophisticated trigger for arevolver, the bullet: caspases.

Typically, two canonical caspase activation pathways have been described (Figure 3). One begins withligand binding to the outer cell-surface receptors known as death receptors; which include tumor necrosis

factor-α and Fas/CD95. As an example, Fas Ligand (FasL) recognition initiates recruitment of FADD protein

to the membrane, mediated by the interaction of its multiple death domains (DD) with Fas; formation of thiscomplex known as DISC (death inducing signaling complex) is followed by recruitment of pro-caspase 8,


which becomes activated upon binding and autocatalytic pro-peptide cleavage. In certain cells, (Type I cells)caspase-8 activation is strong enough to be followed by procaspase-3 processing, activation, death signalamplification and execution. For the rest of the cells -type II- however, an alternative set of proteinsmodulates caspase-8 signal: this is known as the mitochondrial pathway. In the latter the cytoplasmaticprotein Bid is processed by caspase-8 spawning a fragment that relocates in the mitochondrial outer

membrane (MOM), there hetero-oligomerizes with Bax permeabilizing the membrane and inducing extrusionof various signals from the mitochondrion. Bax oligomerization at the mitochondrial membrane is anotherimportant cross point, and is controlled by various pro and anti-apoptotic factors; we will cover this event inmore detail later in this chapter (pg. 64).

Among the polypeptides produced by mitochondria during apoptosis we find AIF (apoptosis inducing factor)and endonuclease G, both implicated in DNA fragmentation and degradation. But most importantlycytochrome C, whose release and cytoplasm accumulation readily follows MOM permeabilization.[3] Latercytochrome C is assembled in presence of ATP, pro-caspase-9 and Apaf-1 (apoptotic protease activatorfactor 1) scaffolding protein into the apoptosome, a large complex where caspase-9 becomes active and

subsequently induces pro-caspase-3 through proteolysis.

The last fail-safe mechanism in the mitochondrial pathway are IAP (Inhibitor of Apoptosis Proteins), anotherimportant family of anti-apoptotic proteins.[4] Their mission consists in blocking caspase-8 and 9 initiatedevents; and for this some family members are able to inhibit proteolytic activity of various cysteineproteases, among others: caspases 3, 7 and 9.

XIAP (X-linked Inhibitor of apoptosis) is a remarkable paradigm of IAP’s caspase inhibitory properties. It iscomposed of a series of three homologous BIR domains: in particular, the second BIR domain (BIR2) isbelieved to be largely responsible for caspase 3 and 7 inhibition. Whereas its inhibitory role towardscaspase 9 has been characterized by Fesik and col.[5] and demonstrated to stem from BIR3 domain; which

binds to the protease and hinders substrate access to the catalytic site. XIAP BIR-2 and BIR-3 domains are

structurally very similar; they both contain three-stranded antiparallel β−sheets and three α-helices with

similar orientations on either domain. Both are held together by a Zn atom, chelated to 3 histidines and acysteine; the main difference between the two consists on the extra C-terminus helix present in BIR-3. Butdespite their similarities, BIR2 and BIR3 inhibition mechanisms towards Caspase-3 and 9 are slightlydifferent. [6] BIR3 exerts its effect on caspase-9 by binding to its N-terminus AVPI consensus motif (Figure4) through a remarkably localized set of interactions, occluding access to the catalytic cysteine. In fact, aswe will see later on, so localized is this interaction that the sole presence of this N-terminus motif is able todisrupt Caspase9/BIR3 interaction. BIR2 on the other hand, operates somewhat similarly by modulating

binding site accessibility, and although it shows certain affinity towards the former tetrapeptidic motif, itscaspase-3 inhibitory activity involves a different binding surface.

In this “rope pulling” game there is still another player: Smac/DIABLO; yet another mitochondrial proteinreleased during cellular suicide with pro-apoptotic activity. Its effects, however, are opposed to XIAP; in factits job consists in capturing the former and liberating caspase-9 active sites.


Figure 4 Caspase-9 inhibition by XIAP-BIR3 domain and subsequent release by SMAC/Diabloprotein.

Structurally, Smac/DIABLO is a dimer of elongated helix bundles and its N-termini peptides share the sametetrapeptidic motif involved in caspase9–BIR3 interaction. As a result, they are able to supplant XIAP in itsinteraction with the cysteine protease (Figure 4) and the sole presence of Smac liberates caspase-9 toperform its grim duty. That is process procaspase-3, transmit and amplify the death signal. [7]

A recurrent observation in a variety of tumors is the presence of large amounts of naturally occurring anti-apoptotic factors; this is a common strategy used by malignant cells to silence suicide signals. In this senseXIAP has been found overproduced in different cancer phenotypes.[8] Naturally, the most obviousalternative to achieve sensitization to apoptosis for such cells is to increase the presence pro-apoptoticcounterparts: i.e. Smac or Smac heptapeptidic surrogate. Indeed, several authors have demonstrated theapoptotic effects of such peptides and validated XIAP as an interesting therapeutic target,[9] and for thisreason a number of academic and non-academic groups have attempted to develop Smac non-peptidicsurrogates.

Given the biological background presented above, it is undeniable that XIAP’s BIR3 domain is very

attractive from a therapeutic perspective. From a methodological standpoint, this system is particularlyenticing, since instead of classic enzyme target this is a protein-protein interaction; which so far has provenrather elusive to traditional drug discovery programs. Still, Smac/XIAP binding interface is remarkably small–a tetra peptide– and may be regarded nearly as a catalytic site. Predictably, it will be much more amenableto small MW inhibitors, and seems as an ideal transition between enzyme-type and protein-proteininteraction targets.

In the following section, we will be using XIAP-BIR3 domain as a test-bench in the implementation ofdifferent selective labeling strategies with interesting features for eventual HTS (High Through-putScreening) programs.


1.1.2 XIAP-BIR3 PRODUCTION AND SYSTEM VALIDATION

The first milestone we have to accomplish if we are to perform any NMR study on XIAP-BIR3 domain isbeing able to produce and purify it in sufficient amounts. For this purpose several constructs were testedand finally XIAP252-340 sequence plus a His-tag chosen for its yield and ease of purification.

Later, we attempted a standard 15N uniform labeling: protein over expression was performed on BL21 hostcells, followed by sonication, cell lysis and Ni affinity purification. NMR samples were obtained after proteinconcentration and dialysis into sample buffer.

Figure 5 15N-1H HSQC experiments for XIAP-BIR3 domain. Left (Black) spectrum corresponding tofree protein, (red) spectrum after addition of 1:1 Smac full-length protein. On the right we can see(green) the effect of N-terminus Smac heptapeptide addition in excess to a sample containing fulllength SMAC/Diablo and BIR3 (1:1) compared to the free (black) spectrum.


Despite XIAP252-348 construct’s solubility and overall exceptional macroscopic behavior, which allowed us toreach up to mM concentrations, we decided to acquire heteronuclear 2D NMR experiments in order toassess fold integrity. As we can see in Figure 5 15N-1H HSQC spectrum quality is excellent (black), and theobserved signal dispersion seems to indicate that the domain is well folded; in fact, the recordedspectroscopic data closely resembles Fesik’s published results.[10]

Next, we decided to examine the recognition properties described in the bibliography for this construct; andfor this reason titrations were performed with two different versions of its natural ligand: Smac. Increasingamounts of full length Smac/Diablo protein were added to our 15N uniformly labeled domain until a 1:1 ratiowas reached (Figure 5, left-red); at this point, most protein signals broaden beyond detection. Given Smac’ssize and shape –it is a highly anisotropic 40 kD homodimer– together with XIAP-BIR3 molecular weight (13kD), the observations could well be a consequence of an enhancement to BIR3’s correlation time linked tothe complexation event. Such apparent size increase has important relaxation implications that only allowthe observation of loop and termini residues.

Interestingly, when Smac’s N-terminus heptapeptide is titrated into the sample an increasing number of

correlations reappear; also, when compared to the free protein spectrum the latter differs considerably(Figure 5, right-black) and nearly almost every correlation has shifted in the presence of heptapeptide.These results can be interpreted as Smac heptapeptide displacing full-length Smac protein to form a 14 kDcomplex, much more amenable to NMR observation and less prone to the previous relaxation issues. Theextent of changes observed for the heptapeptide-complexed-BIR3 is on the other hand remarkable, and iscertainly a consequence of a large rearrangement on the protein surface induced upon peptide binding.

Figure 6 Schematic representations for an NMR assay to detect inhibitors of protein-proteininteractions. Left: HSQC spectrum for small free protein (i.e. XIAP-BIR3). Center: protein-proteincomplex spectrum XIAP-BIR3/SMAC. Right: low MW inhibitor has disrupted Smac/XIAP-BIR3interaction and spectrum corresponds to BIR3-Smac N-terminus surrogate complex.


All in all these data demonstrates that our construct XIAP252-340 is able to recognize both full length andshortened Smac versions, and that as reported in the literature this interaction mainly involves a short SmacN-terminal section. In fact, the experiment depicted in Figure 5 could be envisaged as a general assay toscreen for protein-protein interaction inhibitors. This takes advantage of a frequent nuisance in proteincomplex studies; usually complexes exceed the NMR accessible molecular weight and only individual

complex members can be approached. In such assay only one part of the complex is labeled, and thisremains unobservable in its complexed form as long as no inhibitor is added to the sample. But whenever acomplex disruptor is introduced, the labeled fragment is released from the macromolecular complex and itsHSQC spectrum recovers the appearance of a small protein. (Figure 6)

Lately D’Silva et al. [11] have used this very same idea to screen for inhibitors of MDMD2-p53 interactionand suggest this can also be extended to small protein-protein complexes –which are still observable byNMR- by artificially increasing the complex size with protein tags. Rizo and collaborators [12] have proposeda fairly similar approach to screen for protein-protein interactions; this elegant method uses non-labeledproteins and is able to differentiate between their free and complexed forms by relaying solely on 1H

spectra. There, the authors anticipate enhanced relaxation properties for protein complexes and are able todetect such behavior on the basis of simple spectra addition and subtraction.

1.1.3 W RELEVANCE AND HOT SPOT COMPOSITION

Although protein-protein binding surfaces seem at first sight somewhat unexceptional, being relatively large(some authors estimate around 600-1200 Å2) and flat, the truth is that most display a large degree ofsurface and charge complementarity with their interacting partners. Moreover, contrary to what one couldthink, most often binding energy attained in the protein complex does not come from burial of largehydrophobic surfaces. Instead, as several alanine-scanning studies suggest, [13] a large fraction of thisenergy is contributed by a small number of residues or otherwise named “hot spots”. These key amino acidstend to be situated at the center of binding interfaces, and are usually surrounded by a shell of residuesengaged in relatively unimportant contacts. This shell of amino acids performs an important protective taskin such flat surfaces: occluding water from the hot spot; which allows for efficient burial of key amino acids,

avoids interaction with competing solvent, and improves contacts between complementary hot spotstherefore reducing complex dissociation constant.

Certain amino acid residues, as shown in Table 1, are very common in hot spots, particularly tryptophan(21%), arginine (13%), and tyrosine (12%). On the contrary, residues such as leucine, methionine, serine,threonine, and valine, account for less than 3% of all hot-spot residues. Intriguingly, the nature of thesefavored amino acids is not strictly hydrophobic, and due to the combination of various chemical features

they have the possibility to mediate a variety of interactions (hydrogen bonds, hydrophobic, π−π aromatic

interactions) etc. and even more so within the lowered dielectric environment found in hot spots.

All in all, tryptophan is by far the amino acid with the highest fold increase in its hot spot occurrence, and

this is even more interesting considering it is the least abundant amino acid in proteins. In the same sense,and despite its hydrophobic character, it is as often found in protein cores as it is on surfaces. The samehappens for tyrosine, which as tryptophan, has a large hydrophobic surface and is able to contribute with

aromatic π-interactions and hydrogen bonds. [13, 14] Certainly, hot spot amino acid frequency is not

random and has been shaped throughout evolution to perform a precise task: mediating protein-protein


interaction; and if we are to target such interactions it seems only natural benefit from this privilegedresidues by using them to monitor binding events.

Table 1 Amino acid preferences in protein hot spots according to Bogan et al. [13]. Hot spot aminoacids are those contributing to Gibbs binding energy in 2 kcal/mol or more.

As already mentioned, NMR monitorization of protein-protein interactions faces all the difficulties associatedwith high molecular weight macromolecules, namely relaxation and spectrum complexity issues; some ofthese can be alleviated by selectively introducing stable isotopes and monitoring a few resonances, ratherthan hundreds of them. And what better choice do we have than to carefully place these probes on residues

Residue Total (%) Hot spots (%)Fold enrichment in hot

spots

Arg 218 9.38 29 13.3 2.47

Asn 99 4.26 5 5.05 0.93

Asp 177 7.61 16 9.04 1.67

Cys 3 0.13 0 0 0

Gln 160 6.88 5 3.13 0.58

Glu 220 9.46 8 3.64 0.68

Gly 28 1.2 1 3.57 0.45

His 50 2.15 4 8 1.49

Ile 104 4.47 10 9.62 1.79

Leu 242 10.41 2 0.83 0.01

Lys 143 6.15 9 6.29 1.17

Met 69 2.97 2 2.9 0.54

Phe 166 7.14 5 3.01 0.56

Pro 89 3.83 6 6.74 1.25

Ser 178 7.66 2 1.12 0.21

Thr 131 5.63 2 1.53 0.28

Trp 19 0.82 4 21.05 3.91

Tyr 122 5.25 15 12.3 2.29

Val 107 4.6 0 0 0


found in up to 46% of the hot spots: Trp, Tyr, and Arg. In the following section we will explore selectiveintroduction of NMR observable isotopes on tryptophan, the most relevant of the three.

XIAP/caspase-9 interaction is an excellent example of tryptophan’s ubiquity mediating protein-proteininteractions. In this particular case the interaction surface is limited to a relatively small surface, and most ofthe interaction energy befalls on AVPI tetrapeptidic motif –shared by caspase-9 and SMAC N-terminus- and

on two tryptophan residues in XIAP-BIR3. Liu and collaborators [15] structurally dissected this complex and

observed a set of crucial interactions: D314 forms a salt bridge with peptide alanine (A1) α-amino, while

A1’s methyl group establishes a tight hydrophobic interaction with W310 side chain. Additionally alanine’scarbonyl forms a hydrogen bond with a second tryptophan, W323. The previous tryptophan in BIR3 is alsoinvolved in important hydrophobic contacts with the peptide proline, thanks to tetra peptide extendedconformation. Tryptophans 310 and 323 are thus the foundation of BIR3 interaction with its biologicalpartners: Caspase 9 and Smac. Consequently, introduction of probes at these positions would provide asimple and reliable method to monitor such an important pocket. For the same reason XIAP-BIR3 makes anexcellent test bench in our attempt to devise a selective labeling strategy for tryptophan.

Figure 7 XIAP-BIR3 domain in complex with Smac N-terminus tetra peptide (AVPI). Tryptophanside chains are depicted in sticks.

1.1.4 TRYPTOPHAN SELECTIVE LABELING

At the time of our internship in Dr. Pellechia’s Laboratory, the available methodologies to address specifictryptophan labeling were sparse, and to some extent they were rather general to almost every residue.

Among these, probably the simplest strategy consists on supplying bacterial media with the labeled aminoacid one wishes to introduce the isotope at. However, care must be taken to avoid this being processed bybacteria and scrambling the spin label to metabolically close residues. Isotope dilution can be minimized tosome degree by supplying the rest of amino acids together with the labeled residue, since under theseconditions the bacterial enzymes responsible for amino acid interconversion appear to be low or absent.Still, success is largely dependant on the particular amino acid and where the isotope is introduced; for


instance Asn, Asp, Gln or Glu, being biosynthetically central products, are typically used as precursors toother amino acids and thus prone to suffer from scrambling. While metabolically peripheral amino acidssuch as Trp or His can be labeled with higher isotope purity.[16]

Another general strategy takes advantage of amino acid specific auxotrophic bacterial strains. These strainshave tailored defects on their metabolic machinery that hamper the fabrication of one or more amino acids,

which have to be supplemented in the media for them to survive. There are currently available auxotrophicE. coli strains for every amino acid and although protein yields may suffer from the use of such strains, theycan be useful when one wishes to introduce non-proteinogenic residues. Duewel and collaborators [17, 18]provide several examples for the latter; in their work they introduce fluorinated methionine analogs intophage lysozyme by using a methionine auxotrophic E. coli strain. Similarly, Eichler and collaborators haverecently introduced fluorohistidine analogs into PapD chaperone. [19]

As we have seen above, amino acid production blockade can be introduced at a genetic level; however, onecan achieve the same result by chemically inhibiting certain biosynthetic pathways. Kim and collaboratorsreported successful introduction of fluorinated phenylalanine analogs into E. coli ATPase through

glyphosate-mediated inhibition of aromatic amino acids biosynthesis.[20] There, tryptophan and tyrosinewere supplemented together with 4-fluorophenylalanine to obtain remarkable incorporations for the non-natural amino acid into ATPase.

So far, in all mentioned strategies one needs to supplement isotopically labeled amino acids to the media,which is often an expensive or synthetically difficult choice. A cheaper approach, especially suitable whentrying to label several amino acid types, is the use of common biosynthetic precursors. In this case, ratherthan providing the whole amino acid, a precursor is supplemented; usually the end isotopic incorporationdepends on the particular precursor selected and the reactions this undergoes within the host’s metabolism.There are many successful examples of such strategies in the literature; particularly, Kay and collaborators

have widely used them to incorporate stable isotopes on aliphatic residues, especially on methyl andmethylene groups. Regarding labeling specificity, this can be controlled to some extent with the precursor

choice; for instance 13C pyruvate will introduce labels in Ala, Ile, Val, and Leu methyl carbons, while 13C α-

keto-isovalerate will primarily label Leu, Val residues, and 13C α-keto-butyrate will only affect

isoleucines.[21-23]

Several authors have also used precursor approaches to introduce spin labels on aromatic side chains,these operate similar to methyl and methylene labeling schemes, except that shikimate or chorismate areused instead as common precursors to aromatic residues. In this context, Rajesh and collaborators havevalidated the former scheme by selectively incorporating protons into aromatic side chains in an otherwiseperdeuterated protein background; in this way relevant aromatic nOe restraints were collected for yeast

ubiquitin hydrolase 1. [24]

Despite the apparent variety of methods available to incorporate spin labels on proteins, thus far the abovemethods have something in common: they all require a host to produce the protein; and although this is awell established technology that provides sufficient yields, it often fails to unorthodox labeling schemes. Inthis respect, several recent breakthroughs promise to flatten the way to a wider universe in proteinproduction. Cell free expression systems [25-27] have been proposed, and once some efficiency issues areworked out, they are expected to be much more versatile to the incorporation of non-natural amino acids


and less inclined to isotope dilution; since the supplemented non-natural amino acid and its potential toxiceffects will not compromise the viability of the protein expression system. On the other hand, semi syntheticstrategies like Expressed Protein Ligation (EPL) allow for virtually any chemical modification on a givenprotein section. As a result they have raised considerable interest to produce distinct isotope labelingschemes within segments of the same protein (segmental labeling). [28-30] But probably the “Holy Grail” in

the field is currently being explored by Peter Schultz’s group [31, 32]; they intend to expand the host cellgenetic code in order to incorporate more amino acids than the 20 naturally available.

Keeping in mind our protein production expertise, it seemed wise to concentrate our efforts on selectivelabeling techniques involving E. coli strains as hosts. At the time, the most appealing options to achievetryptophan selective labeling implicated either the use of biosynthetic precursors or some flavor of labeledamino acid in combination with Trp auxotrophic E. coli strain or chemical “knockdown”. Whichever thechoice, it is necessary to understand how tryptophan is produced within E. coli.

1.1.4.1 TRYPTOPHAN BIOSYNTHESIS

In bacteria, fungi and plants the branched pathway leading to tryptophan, tyrosine and phenylalanine [16] isfundamental; not only are these amino acids essential, but also for this is the main route to aromatic ringformation and thus to many secondary metabolites. This proceeds by cyclization of the aliphatic derivative

(2-Dehydro-3-deoxy-D-arabino-heptonate 7-phosphate) and subsequent steps introducing double bonds ina 6-member ring precursor (Figure 8). The first steps produce shikimate, which after three enzymaticreactions is transformed into chorismate, the last common precursor to aromatic amino acids. Next, for Pheand Tyr, biosynthesis proceeds through prephenate as intermediate; while for Trp, chorismate istransformed into anthranilate and subsequently condensed with 5-phospho-D-ribosyl-1-pyrophosphate(PRPP) to form N-(5-Phospho-D-ribosyl)anthranilate. Later, indole-3-glycerol-phosphate synthase willproduce the indole ring from anthranilate’s amino group and two carbons provided by PRPP. Finally, thevery last two steps leading to tryptophan biosynthesis are catalyzed by tryptophan synthase. All in all, itappears that aromatic amino acids are peripheral metabolites, sitting considerably distant from traffickedcrossroads and intermediates, (Figure 8) which is an important consideration if they are to be supplemented

in the bacterial medium as labeling sources, for they will be less prone to scrambling.

In bacteria (i.e. E. coli), tryptophan synthase is a multi-enzymatic complex formed by two different

polypeptidic chains α and β; two copies of each unit are assembled into a (α,β)2-hetero-tetramer, or a

dimer of hetero-dimers. Each chain presents a distinct activity and whereas α unit catalyzes IGP (3-indole-

D-glycerol 3’-phosphate) transformation into indole and G3P (D-glyceraldehyde-3’-phosphate), β subunit is

responsible for the condensation between indole and L-serine using a pyridoxal-5’-phosphate (PLP)cofactor. (See Figure 9)

Tryptophan synthase’s architecture is noteworthy: individual active sites for α and β subunits are connected

through a 25Å tunnel; this is largely hydrophobic and its dimensions allow indole to transit from α to β

subunit active sites. The rational behind this structure is efficiency; since a controlled indole transfer withinthe protein matrix avoids its escape to the cytoplasm and eventually through the bacterial cell wall due to itsneutral charge and hydrophobic nature. Coupled to this transport, a fine network of allosteric mechanisms

orchestrates a precise choreography where the activity of catalytic sites in α and β subunits occurs in

phase. [33]


Figure 8 Aromatic amino acids biosynthesis in E. coli. Extracted from KEGG resources [16, 34] Thelast steps for tryptophan assembly, catalyzed by tryptophan synthase, are shown in more detail inthe bottom chart.

The fact that indole is an intermediate for the last enzymatic steps towards tryptophan is very appealing one.This simple molecule is synthetically very accessible, in fact, it is commercially available in various labeledflavors; additionaly, indole is nearly as “metabolically isolated” as it is the complete amino acid. Bothfeatures are very attractive for labeling purposes as tryptophan side chain precursor. On the other hand,intracellular transport will not likely be an issue, since the very reasons that require the existence of a tunnelin tryptophan synthase will probably play on our side, favoring transit across cell wall.


Figure 9 Schematic representation of Tryptophan synthase (α,β) heterodimer and reactions taking

place within each subunit.

As already mentioned tryptophan synthase machinery undergoes a sophisticated allosteric ball to guide

substrates towards its end products. During the various reaction intermediates α and β catalytic sites are

not always accessible to the bulk solvent, nor activated; as a result there is the possibility that the only

“reactive” indole in β-subunit active site is that which has transited through the 25 Å tunnel; a fact that would

greatly reduce the isotope incorporation efficiency. So, exogenous indole may not be such a straightforwardbiosynthetic precursor, and therefore the hypothesis should first be verified.

1.1.4.2 PROOF OF PRINCIPLE: REVERSE LABELING STRATEGY

In order to test the validity of our hypothesis –usage of indole as an efficient tryptophan side chain

precursor- two labeling schemes where proposed and carried out for XIAP-BIR3 domain.

First, the same 15N uniformly labeled protein batch used for the previous experiments was used to acquirecontrol heteronucler 2D experiments. The second proposed labeling scheme was devised to evaluate indoleincorporation into tryptophan side-chain by means of a reverse labeling strategy. Protein productionproceeded as in the 15N uniformly labeled case, with the difference that prior to IPTG induction weincorporated 14N indole (50 mg/L) into the media. In this way if E. coli assembles tryptophan from thisprecursor instead of building IGP from the start the isotope present in the aromatic side chain will be 14N,thus unobservable to the NMR experiment.

Unfortunately observation of Trp-NHε spin systems for XIAP-BIR3 free domain is not straightforward and,

as portrayed by Fesik and collaborators in their assignment, [10] this is only possible for one tryptophan:

W317, the least exposed of the four, found in what could be considered the domain’s hydrophobic core.


The reason behind the missing crosspeaks is most likely some sort of dynamic phenomenon within XIAP-

BIR3 fold, leading to broad or undetectable correlations for indole NHε in the active site region. Supporting

this evidence, Shin and collaborators have analyzed fold stability for several BIR3 truncated versions anddetermined that minor truncations (XIAP257-347 or XIAP253-356) show enhanced sensitivity to chaotropicreagents; which suggests that conformational exchange may not solely have functional implications, but

could be an artifact arising from XIAP domain dissection.[35, 36] Whatever the reason, addition of Smac N-terminus peptide seems to freeze XIAP-BIR3 conformational sampling and beyond the numerous chemical

shift perturbations induced (Figure 5, right) all Trp-NHε crosspeaks become observable. Consequently, if

one wishes to analyze indole incorporation by means of 15N-1H HSQC and reverse labeling, this will have tobe done in the presence of Smac surrogate peptide.

Under the previous conditions, comparison between 15N-1H HSQCs for control and reverse-labeled proteins(Figure 10) reveals four missing signals for the protein batch grown in 14N indole enriched M9 media. Fesik’s

assigned W317-NHε cross peak [10] is clearly among the missing signals together with other three signals

in the typical Trp-NHε chemical shift region. All in all, it seems clear that 14N-indole has been introduced into

XIAP-BIR3’s four tryptophans. Interestingly, cross peak disappearance or intensity reduction is

circumscribed to the four mentioned resonances and no other peaks seem to be affected, meaning thatscrambling or spin dilution effects are probably unimportant.

Figure 10 Reverse labeling experiment. 15N-1H HSQC section for 15N-uniformly labeled XIAP-BIR3domain (Black) and 14Nε-Trp 15N-U XIAP-BIR3 (Red). Both protein samples prepared at 700 µM in

the presence of two equivalents of Smac N-terminus heptapeptide.


In conclusion, barely for the cost of 15N uniformly labeled protein we have been able to confirm ourhypothesis: indole is an efficient tryptophan side chain precursor and consequently a good means toselectively introduce stable isotopes within this relevant amino acid. We will now be able to attemptincorporation of much more expensive isotopically labeled versions of indole.

1.1.4.3 LABELED PROTEIN PRODUCTION

Once indole incorporation has been demonstrated feasible we decided to pursue incorporation of NMRobservable spin probes into tryptophan side chain. For this purpose we explored labeled indole commercial

sources, which at the time existed in three flavors: 15Nε, 13C4 and 13C2 indole. Due to the important

exchange issues experienced by 15NHε (see previous sections) it seemed inadequate to try 15Nε indole

incorporation, instead we decided to tackle labeling schemes involving 13C with the hope that theseproblems could be alleviated to some extent by the change in nucleus. Hence, two different protein batcheswere produced similarly as in the precious sections, but differing in the supplemented labeled indole. In bothcases, precursor incorporation was assessed using 13C-1H HMQC NMR experiments.

Protein production proceeded smoothly and yields were not affected by addition of either indole to themedia. Judging from heteronuclear 13C-1H experiments (Figure 11) both attempts were equally successful atincorporating indole into Trp side chains; there were however important differences between 13C2 and13C4–Trp spectra.

Figure 11 13C-1H HMQC experiments for 13C-Trp labeled XIAP-BIR3 domain; protein was producedwith 13C2 indole (A) or 13C4 indole (B) as tryptophan biosynthetic precursor. Samples were preparedat 100µM protein concentration, and spectra recorded for the free protein (black), and after

addition of 2 equivalents of Smac N-terminus heptamer (red).

13C2-Trp BIR3 presents three cross peaks in its free spectrum (Figure 11 A), and although this implies that

slow dynamics is less of an issue to this 13C-1H spin system, it is still remarkable to see that two signals arerather broad and that the missing cross peak is only observable at 50˚C (see Figure 14). Smac N-terminus


peptide addition is again the antidote to conformational exchange and induces similar sharpening effects oncross peaks –as reported in the 15N-1H HSQC–, along with broad chemical shifts changes particularly forone of the spin systems.

On the other hand, 13C4-Trp BIR3 HMQC spectrum clearly shows four sharp slightly more disperse crosspeaks; clearly indicating that this labeling scheme is the least prone to suffer from exchange broadening

compared to 15Nε-Trp or 13C2-Trp BIR3. For this reason resonance sharpening upon peptide addition is

hardly observed; nonetheless chemical shift changes do occur, and albeit not as dramatic as for 13C2 theyare mostly limited to one signal.

Altogether, indole has proven a valuable precursor for tryptophan side chain and has allowed us to produceXIAP-BIR3 domain with a variety of probes selectively placed on this important residue. Introduction of 13Cdoes seem to reduce the deleterious exchange effects observed on the indole 15N and allows for an easierligand-binding monitorization as demonstrated with Smac N-terminus heptapeptide peptide with either 13Clabeling. Yet there are several foreseeable drawbacks to these labeling schemes, mainly related withresonance assignment. These issues are intrinsic to labeling selectivity and hence to the lack of observable

isotopes (i.e. backbone 15N and 13C) upon which traditional assignment protocols rely; nevertheless we canenvisage several alternatives to bypass assignment limitations. Probably the most cumbersome assignmentprocedure involves tryptophan mutagenesis, similarly to what has been done for other selectively labeledresidues; [18] other options include nOe detection with reported ligands –i.e. Smac N-terminus peptide- andsubsequent use of published structural data to identify relevant tryptophans. The latter will not always befeasible, but one can also imagine a scheme combining NOESY type experiments with some sort ofbackbone uniform labeling; backbone resonances could be assigned through conventional methods andidentification of backbone-to-side chain nOes could provide assignment information.

In the case of 13C2 labeling, with the available data, it seems we can provide a tentative assignment (see

Figure 11). Resonances 2 and 3 clearly belong to tryptophans 323 and 310 in BIR3 Smac binding site, asthey are the most affected by Smac heptapeptide; thus, the remaining resonances correspond to residues317 and 275. Assuming a similar exchange behavior for 15N and 13C nucleus it seems reasonable toassociate the sharpest heterocorrelation in both experiments to the same residue side-chain: W317 (Figure10-A, crosspeak 1 and Figure 10).

Once we have demonstrated that selective incorporation of NMR observable nuclei into Trp side chain isboth feasible and desirable, during the following sections we will be discussing several potentialapplications, particularly for 13C schemes.

1.1.5 SELECTIVE 13C-TRP LABELING AND ITS USES

1.1.5.1 SCREENING METHODOLOGY

Our first heteronuclear 2D experiments have already shown that both 13C-labeling schemes are quitepowerful at detecting XIAP-BIR3 interactions with its reported ligand Smac N-terminal peptide. So in a way,both schemes retain relevant ligand-binding information while at the same time keeping very simple spectra,both in their absolute number of resonances and extent of ligand-induced changes. Thus classic Chemical

Shift Perturbation (CSP) strategies can be successfully implemented with such labeling schemes to


determine binding constants and binding site identification; moreover the latter can in some instances bedone without the slow dynamics interference often found in protein binding sites.

But beyond traditional CSP approaches, tryptophan’s peculiar features allow us to move a step ahead in ourquest towards spectrum simplification. Tryptophan is relatively sparse in protein sequences, hence signaloverlap is not likely to occur; on the other hand its striking hot spot fold increase [13] certainly anticipates a

relevant function whenever it is found on a protein surface patch. These two properties can be combinedinto an extremely powerful monodimensional Chemical Shift Perturbation assay.

Selectively introduced 13C nuclei can be used to filter 13C2-H or 13C4-H tryptophan spin systems from the restof the protein and background protons; that is to say a monodimensional version of 13C-1H HMQC or HSQCexperiment will produce resonances only for those protons attached to a 13C isotope. Such experiments willlikely have enough resolution to distinguish between the few Trp present in the protein, while at the samebeing sensitive enough to monitor binding events. Indeed, such super-simplified experiments have beenrecorded for our protein, and as shown in Figure 12, 1D 13C filtered 1H spectra for the free and complexed13C-Trp XIAP-BIR3 are clearly distinguishable.

But, if monodimensional HMQC experiments provide similar information as their 2D counterparts, why issimplification such an important issue? “It’s the economy, stupid”.[37] Experiments presented on Figure 12

have been recorded for a 100 µM protein sample within roughly 3 minutes and yet provide nearly as much

information as a 20 fold longer 2D experiment. Such performance improvement clearly exemplifies thethroughput capabilities of 1D 13C filtered 13C decoupled 1H experiments, or conversely, its ability to use farless protein with the same amount of spectrometer time. This should have a huge impact on per-compoundassay cost, a crucial factor if one is to perform a massive screening.

Figure 12 1D 13C filtered 13C decoupled 1H NMR spectra with 13C2-Trp labeled BIR3 in the absence(top trace) and presence (bottom trace) of Smac N-terminus 7mer.

1.1.5.2 INDOLE PROTON DEUTERIUM EXCHANGE AND TEMPERATURE COEFFICIENTS

Exchangeable protons, such as amides or Trp-NHε spin systems, undergo chemical exchange with protic

solvents like water or ethanol. NMR provides several parameters to characterize rate and extent of thischemical reaction; which can be interpreted in terms of solvent exposure for this proton or even draw


structural conclusions, such as hydrogen bonding involvement or its relative position in the protein fold.Exchange rate is typically measured by following proton signal decay with time in the presence ofdeuterated protic solvent, and later these data adjusted to an exponential curve. In particular amide protonexchange, obtained from 15N-1H HSQC or 1H NMR experiments, is a very rich source of information to studyfolding mechanisms, peptide conformational preferences (see chapter 3) and even protein structure.[38]

For XIAP-BIR3’s 15N-1H HSQC direct exchange measurement was impaired to some extent, especially for

tryptophan’s NHε side chains, due to the aforementioned slow protein dynamics and would only be possible

in the presence of Smac N-terminus heptapeptide. Interestingly, as we will see in the following paragraphs,13C2-Trp labeling provides various indirect means to measure this chemical exchange.

To 13C2-Trp, deuterium exchange in the neighboring NHε is seen as a slight change to its carbon chemical

shift and thus an extra cross peak appears for 13C2-1H; this isotopic effect provides a means to detect and

quantify NHε and N2Hε populations and its interconversion rate (see Figure 13).

Figure 13 2D 1H-13C HMQC recorded for samples in 100% D2O buffer (red) and 60% D2O buffer(black); prepared with 0.2 mM 13C2-Trp BIR3 in presence of 0.4 mM Smac N-terminus 7mer.

Addition of 40% H2O to a 13C2-Trp BIR3 / 99% D2O sample followed by prompt HMQC acquisition results inFigure 13’s black experiment. Interestingly, we observe various degrees of exchange for each correlation,

meaning that at acquisition time some of the NHε had not reached exchange equilibrium. In fact, peak

intensity allows us to establish a qualitative exchange rate order for BIR3 tryptophan side chain NHs:

1<2,3<4 (see Figure 13). According to our tentative assignment, the slowest exchanging NHε belongs to

W317, the most buried of the four and as already mentioned in the domain’s hydrophobic core;


unsurprisingly this is followed by residues 323 and 310 both experiencing some degree of burial due tointeraction with Smac’s N-terminus. Finally W275 has the most exposed side chain, a fact that correlateswell with Fesik’s XIAP-BIR3/Smac N-terminus heptapeptide complex structure. [15]

Another useful parameter related, in a way, to exchange rates is temperature coefficient. This has beenwidely used to obtain solvent protection information for conformationally flexible or unstructured peptides as

an alternative to direct exchange rate measurement. For such systems (see chapter 3 for a broader study)exchange is often too fast to be measured; instead, temperature effects on exchangeable protons’ chemicalshift are recorded in strong hydrogen bonding solvents (i.e. water, DMSO or methanol). Hydrogen bonded

protons will experience increased protection to temperature mediated δ changes; in particular, values below

4 ppb/K [39] (Δ chemical shift per Kelvin) correspond to amides considered protected and thus, likely

intervening in an hydrogen bond.

Figure 14 2D 1H-13C HMQC recorded for a 0.2 mM 13C2-W BIR3 sample at different temperatures(red: 303 K, green: 308 K, blue: 313 K, purple: 323 K). Temperature coefficients (ppb/K) have beencalculated using the average chemical shift value (δav); they are represented in brackets next to

each correlation.

As in the exchange rate case, direct measurement of Trp-NHε temperature coefficient would not be very

informative due to our inability to observe most of these NHs, particularly for BIR3 free form. Fortunately,

temperature effects on NHε chemical shift are disseminated via through-bond effects to the vicinal 13C2-1H,

thus a similar chemical shift versus temperature plot is obtained for the C-H system. This exercise has been


carried out for XIAP-BIR3 (Figure 14) and the averaged chemical shift slope, upon temperature increase,calculated for the available 13C2-1H cross peaks.

Temperature coefficient interpretation should in principle be analogous to that of amide protons; however,

due to the lack of literature with respect to Trp 13C2-1H or even NHε coefficients, it is difficult to establish a

reliable threshold below which a tryptophan side chain can be considered protected from solvent. Therefore,

we are left with their relative solvent protection within XIAP-BIR3. In this context, W317 (Figure 14,crosspeak 1) would not be involved in any hydrogen bond, especially when compared with the other twoobservable signals (2, 4) -one belonging to W270 and another to a Trp in Smac’s binding site-, which are farless affected by temperature. According to XIAP-BIR3 structure [10], only W323 and W270 would be able tohydrogen bond neighboring carbonyl functions, in agreement with our measured data; however, this mustbe taken with a pinch of salt since Fesik’s NMR structure may not be completely reliable when it comes toside chain conformations.

1.1.5.3 OTHER POTENTIAL USES

So far, our proposed labeling schemes have been tested in their ability to identify and characterize smallprotein-ligand complexes; but eventually, the size of such systems could be expanded through acombination of selective labeling schemes and TROSY type experiments. TROSY was first introduced in aseminal work by Pervushin and collaborators [40]; there the authors achieved mutual compensation of CSA

(chemical shift anisotropy) and dipole-dipole relaxation, yielding improved sensitivity and spectral resolutionfor 15N-1H spin systems. Lately, similar experiments have also been developed by Pervushin andcollaborators to tackle aromatic 13C-1H systems.[41, 42] It is in this respect that we believe there could beinteresting implementations for our Trp selective labeling schemes, especially 13C2-Trp; which due to itsrelative isolation, could probably deal with very large macromolecules or complexes in combination withperdeuteration.

Dynamic data is yet another type of information one could obtain from tryptophan side chain labeling; inparticular, measurement of various relaxation parameters could provide an interesting portrait of protein hotspots dynamics, [43] which is arising more and more interest in the understanding of protein function.

Another active field where our methodology could have an impact is protein folding. Tryptophan, when inprotein cores, usually performs relevant duties during folding reactions or fold stabilization; for this reasonseveral authors have used tryptophan probes to monitor such events. [44] Our labeling schemes could beused as an alternative to perform such studies.

1.1.6 SELECTIVE LABELING WITH 19F-TRP

Throughout the previous sections, our understanding of tryptophan biosynthesis has allowed us to introducevarious NMR observable isotopes at its very side chain. Yet isotopes have been limited to 13C and 15N,which is in essence an enrichment exercise for certain protein positions in order to enhance their sensitivityto NMR experiments. And, while this approach is very desirable in order to avoid disturbing protein fold andactivity, it still relays on 15N-1H or 13C-1H detection that has to compete with an important signal backgroundin the experiment, mainly arising from solvent protons and natural abundance 13C.


Recently, 19F utilization has attracted increasing attention and been proposed to deal with such issuesparticularly in several HTS NMR experiments. [45] This nucleus displays some especially compellingfeatures: first its sensitivity, stemming from 19F high natural abundance and its large gyromagnetic constant

–roughly 80% of γΗ –; also, its wide ppm range and chemical shift sensitivity to subtle changes in the

electronic environment have proved extremely useful. Finally, fluorine is not a common atom in

macromolecules or even secondary metabolites and will very likely solve NMR background issues, thusreducing protein requirements and overall assay cost.

19F has traditionally been regarded as decent 1H isoster in medicinal chemistry due to its slightly larger Vander Waals radius; but despite of their size similarity, the two nuclei are rather different in their electronicproperties. In addition, fluorine has demonstrated to enhance lipophilicity; which on the other hand may beinteresting from a medicinal chemist perspective to modulate a compound’s protein-binding properties.[46]For these reasons, care should be taken when introducing 19F into proteins as fold distortions may occur; infact, Duewel and collaborators have reported such phenomena when fluorinated methionine analogs are

built into bacteriophage λ lysozyme (LaL).[17]

No matter how close we perceive 1H and 19F to be, the truth is that E. coli –or any of the frequently usedhosts– does distinguish between natural amino acid and a fluorinated analogs; as a consequence, proteinbiosynthesis yields are greatly diminished with such amino acids either because cellular machineryrecognizes them as exogenous or due to some intrinsic toxicity. Several alternatives to increaseincorporation yields have been already mentioned, amongst the most straightforward we find: chemicalinhibition of amino acid biosynthesis and the use of auxotrophic hosts.

Kim and collaborators successfully demonstrated the former approach for aromatic amino acids [20] bychemically blocking shikimate pathway via 5-enolpyruvylshikimic acid-3-phosphate synthase inhibition. Inthe presence of glyphosate (1 g/L), incorporation of fluorinated aromatic residues into proteins proceeded

much more efficiently, though not completely, than by simply supplementing the exogenous amino acid tothe bacterial media. In the following section we will be exploring a similar approach, only slightly morespecific, to introduce 19F-Trp analogs using our model protein XIAP-BIR3.

1.1.6.1 TRYPTOPHAN BIOSYNTHESIS INHIBITION: PROOF OF PRINCIPLE

Indole-3-acrylic acid (IAA) (Figure 16) has long been known to lead to tryptophan starvation in bacteriahampering normal growth rates; and according to several authors, its mode of action combines severalsimultaneous and opposing targets. Even though it has the ability to prevent Trp repressor to work, and thusenhances expression of enzymes in tryptophan biosynthesis, it also inhibits tryptophanyl-tRNA synthasepreventing Trp incorporation in polypeptide chains [47-50]. IAA has also been proposed as an inhibitor of

Tryptophan synthase [51] and is believed to interfere with the formation of the α -aminoacrylate

intermediate; necessary to initiate the second reaction stage in Tryptophan synthase β-subunit and

therefore inducing indole accumulation in bacteria (Figure 9).


Figure 15 E. coli proliferation experiment in the presence of Indole-Acrylic Acid. () Culture mediasupplemented with IAA. () Culture media supplemented with IAA and L-tryptophan.

From the above literature, it seems that whatever the mechanism behind tryptophan starvation, this islimited to this residue and that production of any other amino acid within bacteria will not be affected.However, and given the apparently contradictory mechanisms by which indole-acrylic acid exerts its effect,we decided to perform a proliferation experiment on E. coli under IAA effects. Two E. coli cultures weregrown independently in M9 minimal media; the first culture was solely supplemented with IAA, while in the

second tryptophan was also included. Bacterial growth was monitored with time using standard opticaldensity measurement (at 600 nm wavelength) and the result plotted as in Figure 15.

IAA effects on E. coli growth are clear and agree with the previous literature: at a concentration of 50 mg/ml,indole acrylic acid seems to induce tryptophan starvation on bacteria and as a result proliferation arrest; cellgrowth is on the other hand restored by supplying the amino acid to the media, clearly indicating aninhibitory effect on the production of this amino acid.

1.1.6.2 19F5-TRP XIAP-BIR3 PRODUCTION

Once we demonstrated that IAA induced tryptophan biosynthesis inhibition, and conditions for suchinhibition were optimized, we decided to pursue protein production where L-tryptophan was supplanted by afluorinated analog: 5-fluorotrypotphan. (Figure 16)


Figure 16 Chemical structures. (Left) 5-fluorotryptophan and (Right) indoleacrylic acid.

BL21 E. coli cells, containing XIAP-BIR3 plasmid, where grown in M9 minimal and prior to induction bothIAA and fluorinated tryptophan analog were added to the culture. Induction, cell lysis and purificationproceeded as described for previous protein batches. Although protein yield was reduced to 50% -comparedto unlabeled protein expression-, the sole production of protein was in itself encouraging given the presenceof IAA, and is certainly an evidence for either 5-fluoroTrp reversing IAA inhibitory effect or the formerfunctioning as a tryptophan substitute. On the other hand, a protein yield drop was not completelyunexpected considering the reported 5-fluoroTrp toxic effects on bacteria [52].

In order to evaluate the extent of fluorotryptophan incorporation, and prior to 19F NMR experiments, we

decided to perform protein MALDI-TOF MS analysis. Both for XIAP-BIR3 produced in the presence of 5-fluoroTrp and IAA inhibitor, and for an unlabelled control protein batch.

Figure 17 MALDI TOF mass spectra of 5-fluoroTrp BIR3 (a) and of unlabeled BIR3 (b)


MALDI TOF spectrum for the non-labeled XIAP-BIR3 domain displays a major peak at 13106.6 Da. Whilefor 5-fluoroTrp supplemented sample two peaks are obtained, the most intense of them at 13178.1 m/z andthe other at 13158.5 m/z with roughly a 40% intensity. Fluorine substitution of all H5-Trp would translate intoan 18 units difference with respect to wild-type protein; interestingly MW difference between the two spectramajor peaks is 72 a.m.u, the exact figure one would expect for a 4x19F-Trp substitution. Incomplete 19F-Trp

incorporation could explain the lighter peak at 52 a.m.u, with nearly the expected mass of a three fold 19F-Trp substitutions; the error may be attributed to inherent resolution of linear TOF acquisition.

From the data above, it would seem that fluorinated tryptophan incorporation has been quite successful andat least from a qualitative perspective this has occurred for 90% of Trp residues ((0.6*1+0.4*0.75)*100). Thisincorporation yield is far superior to the strategy where solely fluorinated tryptophan is supplemented [53],but remarkably our approach is also superior to the use of glyphosate as inhibitor with a reported 80%incorporation. [20]

The ultimate 19F-tryptophan incorporation tests were a series of 19F NMR experiments. (Figure 18) As in theMS assays fluorine presence was confirmed; in particular for the free BIR3 domain three signals are clearly

observable, two of which being rather broad. The situation is analogous to that of 13C2-Trp selective labelingand is possibly due to the same protein dynamics phenomenon; again, addition of Smac N-terminus peptidefreezes this conformational rearrangement and induces considerable 19F resonance sharpening, along with

important δ perturbations. Two out of the three original signals experience important shifts (0.8 and 0.5 ppm

respectively) and a fourth signal appears upon peptide addition.

The observed spectrum changes for 5-fluoroTrp BIR3 are consistent with those observed throughout thischapter and could hence be interpreted similarly. (Figure 18) Analogously, a tentative assignment could beattempted; but in order to obtain a reliable identification one should perform a thorough site-directedmutagenesis study.

Figure 18 1D 19F NMR spectrum of 5-fluoroTrp XIAP-BIR3 without (A) and with two Smac N-terminus peptide equivalents (B). Spectra were collected at 25 ºC with 0.1 mM protein sample of19F-Trp in H2O/D2O (90:10) phosphate buffer (40 mM, pH 7.5). Ribbon structure of the free andcomplexed XIAP-BIR3 is depicted on top of each 19F spectrum.


As expected, fluorine is an extraordinarily sensitive probe and in our case is able to monitor Smac/BIR3interaction with a performance comparable to that of previous 13C-Trp schemes; and although it does notcarry as much structural information as 13C2-Trp its is superior in terms of signal dispersion and background.For this reasons 19F-Trp labeling should have an important impact on NMR screening methodologies,reporting most of the advantages related to selective isotope incorporation on tryptophan without any of the

deleterious background effects. Solvent or other background signals often determine the lower sensitivitythresholds for many such experiments; thus, 19F acquisition will most likely reduce protein requirements andoverall assay costs.

Beyond HTS applications, 19F labeling could also contribute to folding and protein dynamics studies;although the range of macromolecules amenable to such studies will probably be somewhat limited due tofluorine intrinsic relaxation properties. Fluorine nuclei in proteins relax predominantly by means of dipole-dipole interactions with surrounding protons and by the chemical shift anisotropy mechanism; bothmechanisms are particularly enhanced with correlation time and magnetic field hampering the study of largemacromolecules or complexes. [53]

In summary, we have devised a novel approach to selectively incorporate 19F in protein tryptophan sidechains, this has demonstrated easier and more efficient than previously reported methods. Our methodologyis cheaper than the use of glyphosate (250 Euros/5 g) and requires minor amounts of inexpensive IAAreporting good protein yields. On the other hand, this scheme is able to successfully monitor complexationof our protein with a short peptide and is expected to have a wide range of applications in HTS, proteinfolding and dynamics.


1.2 NMR SCREENING STRATEGIES APPLIED TO BCL-XL

1.2.1 BCL-XL BIOLOGICAL ROLE AND INHIBITION RELEVANCE

From the previous overview on apoptotic signaling pathway, it is clear that a crucial checkpoint to unleashcaspase activity is established on mitochondria.[2, 3] In particular, release of proapoptotic proteins such ascytochrome c is controlled through a sophisticated mechanism by which certain Bcl2 protein family

members heterodimerize and translocate to the mitochondria outer membrane (MOM). MOM becomespermeable to the release of cytochrome c and the death signal is transmitted downstream via assemblageof the apoptosome. (Figure 19)

Bax and Bak are the most relevant proapoptotic members in Bcl2 family, those that ultimately dimerize onmitochondria membrane triggering cytochrome c release. However, prior to heterodimerization, aconformational change must occur on these proteins for them to expose important hydrophobic helicaldomain (BH3 helix). This conformational change is induced by a variety of signaling proteins that also havethis BH3 helix in common (‘BH3-only’ molecules) and are generated upon death stimuli. For instance, Bidtruncated peptide is generated upon caspase-8 activation, while other “Bid-like” molecules (NOXA, PUMA)are produced as a response to DNA damage or cytoskeleton disruption through transcriptional control.

On the other hand, several members of the Bcl-2 family assume the opposite role and perform anantiapoptotic function by desensitizing cells to death stimuli. Bcl-Xl and Bcl-2 belong to this group and bothdisplay homology to the four BH domains in Bak, interestingly they exhibit affinity towards BH3 domains.

Figure 19 Model for the role of Bcl-2 family in apoptosis. (See text for more details) Extracted fromBrenner and col. [2]


This exquisite balance between pro and anti-apoptotic Bcl-2 family members determines mitochondrial pro-apoptotic proteins liberation and ultimately the cellular fate; not surprisingly miss-regulation of suchequilibrium has been identified as an important contributor to tumorigenesi and tumor resistance. In fact, anumber of the mentioned desensitizing (Bcl-XL, Bcl-2) proteins are found over-expressed in many cancersand operate by capturing pro-apoptotic members of the family via BH3 domains to silence their signals.[9]

The simplest way to neutralize this decreased sensitivity to pro-apoptotic signals consists in increasing theamount of stimuli; and indeed, presence of Bak BH3 peptide has been reported to induce on its owncaspase mediated apoptosis on HeLa cells; thanks to its ability to occupy Bcl-2 and Bcl-XL hydrophobic cleftliberating Bak and Bax. Such evidences validate both BCL2 and BCL-XL as cancer therapeutic targets,even more to the light of BCL-XL/BIR3 complex structure solved by Fesik and collaborators; which hasfostered various screening and rational design efforts to build non-peptidic BH3 analogs capable ofpreventing BCL-XL’s anti-apoptotic role.

Figure 20 Bcl-XL (ribbon) in complex with BH3 Bak 16mer peptide (sticks and dotted surface).[54]

BH3 peptide binds to Bcl-XL through a cleft situated between BH1, BH2 and BH3 Bcl-XL helix domains, thisinteraction is mainly dominated by hydrophobic contacts although alanine scanning [54] has also revealedseveral important electrostatic and hydrogen bonding interactions. (Figure 20) Comparison between free

and BH3 complexed structures highlights significant conformational changes that occur upon peptidebinding, especially involving exposure of the large binding hydrophobic patch.

Degterev and collaborators [55, 56] have also reported similar rearrangements for Bcl-XL in the presence ofseveral small molecular weight ligands; they would apparently also be capable of stabilizing the former“open cleft” conformation. The previous small molecules were identified in a fluorescence polarization assayfor their ability to compete with BH3 peptide binding to Bcl-XL; later they demonstrated to induce in vivoapoptosis on a tumor cell panel through a mechanism consistent with Bcl-XL heterodimerization inhibition.


[56] Initial attempts to characterize by NMR the complex between Bcl-XL and BH3I, one of the moleculesreported by Degterev, were unsuccessful due to its affinity regime –fast or intermediate exchange- that oftentranslates in a limited number of experimental restraints. Another difficulty was related to the mentioned Bcl-XL plasticity; large conformational changes were observed upon BH3I-1 and BH3I-2 binding, thus traditionalChemical Shift Perturbation (CSP) analyses would not provide a reliable picture of their binding sites.

Instead, a differential CSP [55, 57] strategy was applied, and a battery of 15N-1H HSQC titrations performedfor various BH3I-1 and BH3I-2 close analogs (see Figure 21). The differences in their protein perturbationprofiles were introduced as restraints into an in silico docking protocol to obtain complex structures. Bothtypes of inhibitors seem to be placed in the same BH3 peptide groove and although they overlap to someextent, their binding mode is slightly different. According to the authors, BH3I-2 type inhibitors would cover alonger stretch of the cleft establishing energetically relevant interactions with various aromatic and aliphaticresidues.

Nonetheless, this has not been the only successful story in trying to obtain BH3 peptide surrogates.[58, 59]Huang et al. reported the first small MW Bcl-XL inhibitor as a result of a computational screening and similar

strategy proved equally fruitful to Wang et al.[60, 61] Other authors have used so-called “rational design”and obtained helix mimics to perform Bak-BH3 peptide job. In this sense, Hamilton and collaborators [62-64]have successfully employed oligo-amide foldamers and other scaffolds to spatially locate chemical groupsas found in the inhibitor peptide, later they have tested their ability to disrupt Bcl-Xl- Bak interaction both incellular and biophysical assays. Similarly, Schepartz and collaborators used the “protein graphting”approach with the same end; there, instead of foldamers a restricted peptidic scaffold was chemicallydecorated to mimic BH3 peptide.[65]

Figure 21 Chemical structures for BH3I-1 and BH3I-2 inhibitors [56] and analogs used forDifferential CSP strategy. Measured affinities for each analog are also presented.


All in all, Bcl-XL BH3 binding surface is not the typical flat protein-protein interface; instead, it is a flexibleand quite promiscuous hydrophobic groove that has proved targetable to organic molecules, peptides andmini-proteins. The previous relatively successful stories are encouraging and the nature of protein surface,together with the ligands identified so far, suggest that Bcl-XL may not only be an interesting biologicaltarget but one suitable for fragment based screening. In fact, a recent work by Hajduk and collaborators

mention Bcl-Xl as a particularly druggable system to the light of its high success rate in several screeningefforts. [66]

1.2.2 BCL-XL

15N uniformly labeled Bcl-XL was produced as described by Fesik et al. [67] and supplied by courtesy ofReed’s laboratory at the Burnham Institute. The construct was not full length Bcl-XL, since the putativetransmembrane domain (residues 45 through 84,) was removed to improve solubility and overall behavior.Nonetheless, and according to the published data, the polypeptidic chain retains its anti-apoptotic propertiesin vivo and is able to recognize BH3-Bak peptide. The latter was demonstrated by means of NMR 15N-1HHSQC experiments; these were recorded on an appropriately labeled Bcl-XL sample with and without Bak-BH3 peptide.

Figure 22 15N-1H HSQC for 15N uniformly labeled Bcl-XL. Black spectrum was recorded for a 300µM protein sample; red spectrum was acquired after addition of one equivalent of Bak-BH3.


Heterocorrelation spectrum for the free protein confirms the construct’s large helical content, as expectedfrom the published spectroscopic data. Later, we observed that Bak-BH3 peptide addition has a widespreadeffect on the protein spectrum inducing numerous shifts, which are also consistent with its published affinityand binding mode. (Figure 22)

In conclusion, the protein target has been successfully validated both in terms of fold and recognition

properties; unfortunately due to important overlapping issues –understandably, given the helical proteinnature- assignment of ligand induced perturbations cannot be carried out thoroughly using published proteinassignments. This will be an important limitation during the upcoming sections.

1.2.3 FRAGMENT LEAD DISCOVERY AND LIBRARY COMPOSITION

As already mentioned, NMR experiments can be wisely incorporated into fragment lead discoveryapproaches taking full advantage of their wide range of detected affinity ranges in ligands, throughputproperties and structural information; but more importantly, dodging the particular pitfalls of eachexperiment. [68, 69]

This idea can be implemented in various ways and include other techniques than NMR, however during theinternship in Dr Pellecchia’s laboratory we applied a strategy that relied solely on NMR. Experiments wereorganized as follows: in a first stage the throughput capabilities of cross-saturation experiments such as

STD or WaterLOGSY were used to screen the bulk of a fragment chemical library. Once positive bindersare identified, the second part of the process consisted on their binding validation and the extraction ofstructural information to be used in the design of tighter hit analogs, namely by means of 15N-1Hheteronuclear and/or competition STD or WaterLOGSY experiments, among others.

Primary fragment screenings are often successful for a wide range of protein targets; in fact, it is notuncommon to identify various hits from a single screening round. This is especially desirable when theirbinding modes are slightly different, as it is a good source of inspiration to improve their binding. In the bestcases, primary screening may even provide hits with different binding modes or targeting distinct proteinsites; which for design purposes is a very valuable scenario as a fragment-linking strategy can be applied.Prarticularly when these sites are adjacent and positive compounds sit next to each other on the protein

surface.

In order to diagnose the above situations one can map the binding sites for the various positive fragmentsby means of 15N-1H HSQC CSP studies, alternatively one can also establish whether compounds share thesame binding site using competition cross-saturation experiments (STD or WaterLOGSY). But probably themost informative experiment, not only diagnostic wise but for design purposes, is ILOE.[70] In thisexperiment inter-ligand nOes, mediated by formation of the ternary complex, are detected and can be usedto determine the relative orientation and distance between the two fragments on the protein surface. Later,this information is used to design a convenient linker that will connect the two fragments yielding a dualligand, hopefully with enhanced binding. Interestingly this information is extracted without the need of

labeled protein or assignment and with the cost advantages of ligand based experiments.

ILOE based fragment linking strategy as outlined above [71] offers obvious benefits over schemes relyingheavily on receptor-based experiments, for which structural analysis requires prior knowledge on proteinassignment and structure; moreover binding site mapping is often fuzzy and unreliable due to distal


conformational changes (see chapter 1.1.4.2 and chapter 2). For these reasons, whenever differentfragments are identified in a primary screening, those hits with vicinal binding sites will be prioritized and achemical linkage devised on the basis of ILOE experiment.

With this general strategy in mind, an all-purpose fragment library was designed and the particular chemicalcompounds selected, purchased and assembled for screening. The general criteria used to select the

chemicals conforming a fragment library have already been mentioned (“rule of three”), in our case wefollowed somewhat more stringent guidelines by setting molecular weight and lipophilicity upper thresholdsat 250 Daltons and 2 for LogP respectively. Molecules were selected from a family of scaffolds similar tothose employed in the SHAPES approach; [72, 73] the underlying idea is to select compounds withchemical features commonly found in commercial drugs and filter them according to our molecular weightand lipophilicity thresholds.

Roughly 4-5 compounds per scaffold (Figure 23) were selected and purchased. In terms of logistics, eachchemical was given a code, weighted and dissolved in d6-DMSO at 100 mM as concentrated stocks. Later a1H NMR spectrum was acquired for each compound under typical NMR assay conditions: 0.5 mM

compound in a 50 mM NaCl, and 25 mM phosphate buffer at pH 7. This typical library set up is veryconvenient, as sample preparation and screening are easily performed. On the other hand, 1H NMRexperiments performed for all compounds are very useful as reference spectra, but also in order to verifycompound integrity and solubility under experimental conditions; albeit compound solubility should havebeen addressed to some extent by LogP filter.

Figure 23 Molecular frameworks used for SHAPES library selection. [72] Attachment points for sidechains are indicated by single electrons or lone pairs. ‘X’ represents a C, N, O or S atom.


1.2.4 SCREENING

Bcl-XL is a widely accepted therapeutic target and in our hands has demonstrated to recognize its naturalligand suggesting this could be a suitable protein for fragment based lead discovery. Once we have set upan all-purpose fragment library, observing well-established guidelines, we will now report the results of itsprimary screening using WaterLOGSY experiments.

Figure 24 Screening results for sample containing compounds: 1E11, 1E12, 2A1-2A8. A)Watergate 1H experiment for compound mixture in the presence of Bcl-XL. B) PositiveWaterLOGSY experiment. C) control 1H NMR experiments for compound 2A4 and D) compound2A6.

Samples containing 10 µM Bcl-XL together with mixtures of ten compounds (0.5 mM per compound, 25 mM

pH 7 phosphate buffer and 50 mM NaCl) were prepared for all chemicals in our fragment library. Next, aWaterLOGSY experiment was performed on every sample and one sample identified to contain positivecompounds; this mixture was made up by codename compounds: 1E11, 1E12, 2A1-2A8. Deconvolutionproceeded by comparing positive WaterLOGSY experiment with 1H NMR experiments for free compounds;


and as we see in Figure 24 cross-saturation signals in the positive experiment belong to compounds 2A4and 2A6; 4-benzyl-2-nitrophenol and (Z)-1,7-dimethyl-5-phenyl-1H-benzo[e][1,4]diazepin-2(3H)-onerespectively.

Later, we validated 2A4 and 2A6 binding by performing the same WaterLOGSY experiment for theindividual compounds in the presence of Bcl-XL, and although we observed residual cross-saturation

signals for 2A6 the effect was far greater for benzyl-nitrophenol (2A4) compound.

Altogether, primary screening experiments indicate that both 2A4 and 2A6 bind to Bcl-XL although the latterprobably does so with lesser affinity. Interestingly, a BH3I-1 analog ((E)-2-(5-(4-fluorobenzylidene)-4-oxo-2-thioxothiazolidin-3-yl)-3-methylbutanoic acid Figure 25) also experiences cross-saturation a similar

WaterLOGSY experiment upon Bcl-XL addition. This is not surprising given the µM affinities reported by

Degterev and collaborators;[56] these are within the typically detectable affinity range for cross-saturationexperiments. The above primary screening was quite profitable, two novel Bcl-XL binding scaffolds wereidentified and a BH3-I analog has also been evaluated with positive results.

Figure 25 BH3I-1 fluorinated analog: ((E)-2-(5-(4-fluorobenzylidene)-4-oxo-2-thioxothiazolidin-3-yl)-3-methylbutanoic acid.

1.2.5 SCREENING FOLLOW UP

A closer inspection to the positive fragments identified in the previous section (2A4 and 2A6) reveals thatthey both share a common diphenylmethane motif. This is not surprising, and as a matter of factdiphenylmethane and biphenyl scaffolds are the most recurrent hits in NMR screening programs; someauthors suggest this is the result of protein binding sites intrinsic properties.[73] In the current section we will

examine in detail binding for 2A4, the simplest of the two hits and the one displaying the better cross-saturation effect, so as to extract information to eventually improve binding.

First we performed a classic CSP study to validate 2A4 binding. For this purpose successive 2A4 additionswere performed on a 15N uniformly labeled Bcl-XL and monitored with 15N-1H HSQC experiments. As aresult, a large number of cross-peaks become affected upon addition of diphenylmethane compound and,although most of these changes occur in the fast exchange regime, some signals experience significantbroadening. However, accurate binding site mapping is not possible due to the large number of affectedresidues and the difficulty to identify resonances with the published assignment. [55] The latter probably asa result of Bcl-XL’s crowded spectrum, typical of highly helical proteins. Even so, some amides could be

identified including A168 and R165 in Bcl-XL-BH1 domain; both are remarkably affected by 2A4 additiondespite of their distance to the active groove. (See Figure 26) This is consistent with a large rearrangement


occurring on BH3 BH2 and BH1 domains to expose Bcl-XL’s hydrophobic cleft; a phenomenon that appearsboth upon addition of Bak-BH3 peptide, but also in response to the small inhibitors reported by Degterevand collaborators. [56]

A direct consequence of the poor assignment is that our understanding on 2A4 binding is rather limited andfor now we will not be able to determine whether 2A4 binding-mode resembles that of Degterev’s BH3I-1

compound, BH3I-2 or none of the former. Still, beyond detection of the binding event, 15N-1H HSQC titrationprovides some very interesting piece of information: 2A4’s binding constant. [69]

In order to estimate 2A4 binding constant one must first define an averaged chemical shift difference for

amide cross peaks: ΔδHNav. (See Equation 3) This value takes into account both 1H and 15N chemical shifts

(weighted according to the spectral window of each nucleus) and allows us to reliably quantify chemical shiftperturbations on the 2D heteronuclear experiment.

Figure 26 Bcl-XL 15N-1H HSQC experiments in the presence of increasing amounts of 2A4: (black)free Bcl-XL, (red) 1 equivalent 2A4 (blue) 5 eq. (green) 10 eq. (magenta) 15eq. (cyan) 20 eq.


Figure 27 Non-linear curve fitting of ΔδHNav(ppm) versus compound concentration (mM) for an

amide peak shifted upon 2A4 addition. ΔδH Nav for each HSQC experiment are calculated

according to Equation 3 and fitted to the presented equation with the resulting KD, A and Rcorrelation.

€

ΔδHNav = 12 ΔδH( )2 + 1

5ΔδN( )2( )

Equation 3 Averaged chemical shift difference for 15N-1H heteronuclear correlations.

€

PL[ ] =1 2 KD + P[ ]0 + L[ ]0( ) − 1 4 KD + P[ ]0 + L[ ]0( )2− L[ ]0 P[ ]0

PL[ ] = A ⋅ ΔδHNav

P[ ]0 ⋅ A = ΔδHNav[ ]boundEquation 4 Used in non-linear curve fitting for ΔδHNav vs ligand concentration plot.

Later, one should determine this averaged chemical shift difference with respect to the free protein spectrum

for all titration experiments and desired amides. Δ δHNav values would be plotted versus ligand

concentration and non-linearly fitted to Equation 4 in order to obtain KD.

This exercise was successfully carried out for 2A4 titration and the resulting estimated dissociation constantis 1.6 mM, with a very good correlation coefficient. (See Figure 27) As expected from the observations in15N-1H HSQC and WaterLOGSY experiments, 2A4 is a weak binder in the high-µM to low-mM range.

1.2.5.1 COMPETITION EXPERIMENT AND INTERLIGAND NOE

Due to the limited structural information regarding 2A4 binding on Bcl-XL, it remains unclear whether thiscompound shares the same binding site as BH3I-1 type compounds. For this reason a series of


WaterLOGSY experiments were performed in the presence of both compounds: diphenylmethanecompound (2A4) and Degterev’s fluorinated analog (Figure 25).

Interestingly, addition of the slightly tighter binder (BH3I-1) does not affect cross-saturation effects observedon 2A4, in fact, such effects were observable for both compounds overall suggesting that they do notcompete for the same site on the protein surface. This was later confirmed by acquiring NOESY

experiments for the ternary mixture. The resulting experiment shows intense negative nOes for bothcompounds, which is a typical behavior for high MW molecules and a clear confirmation that bothcompounds are undergoing exchange between bound and free states. The presence of such transferrednOes (trNOE) is very useful to determine bioactive conformations for small flexible molecules, and someauthors recommend these experiments as a reliable screening method.[69] But more interestingly, besidesintra-molecular nOes we can also identify interligand cross peaks between a 2A4 aromatic and aliphaticprotons in BH3I-1 analog; these are also negative nOes and correspond to contacts established on theternary complex (ILOEs).

ILOEs are very informative in the context of a ternary complex as they demonstrate that two given

molecules target different protein binding sites and, should it be the case, one could rule out the possibilityof protein-mediated spin-diffusion effects by using QUIET-NOESY experiment. [74] Furthermore, they affordinformation regarding the relative spatial orientation of the ligands sitting on adjacent sites; in our case 2A4and BH3I-1 would be facing each other with their methyl and orto-phenolic protons respectively as depictedin Figure 28.

Figure 28 NOESY (τm=600 ms) experiment section for 20 µM Bcl-XL sample in the presence of 0.5

µM 2A4 and BH3I-1 fluorinated analog. Relevant iLOES are depicted on spectrum and chemical

structures (dotted line).


1.2.6 BILIGAND DESIGN AND FUTURE PROSPECTS

Once the relative positions of Bcl-XL binding fragments have been established one has to devise a way tolink them; we will consider various possibilities within this last section. But prior to the design, from apractical perspective, it is important to remind us that chemistry involved in fragment linking should be keptas simple as possible, since a robust chemistry will allow us to explore important parameters such as linkerlength and chemical nature, before addressing its thorough optimization in later stages.

Following this “keep it simple motto”, we should look at 2A4 and BH3-I analogs containing, close to the

resonances experiencing iLOEs, chemical functions that can be used in an eventual chemical linkage.Functions such as COOH, OH and NH2 are particularly interesting; they can easily produce esters andamides under mild conditions, with a well-established scalable chemistry, thanks to the availability of solidphase synthesis strategies and resin-attached reagents.

Analogs with anchoring chemical functions should in principle be tested for their ability to recognize Bcl-XLwith the same WaterLOGSY/NOESY experiments, and once binding analogs have been identified a suitablelinker is chosen according to the chemical functions involved. Next, in order to determine the best linkerlength one could use information from in-silico docking methods or ILOE build up curve information;however, a more pragmatic approach consists in the synthesis of a small collection of dual ligands with

variable length linkers.

Figure 29 Proposed design for a Bcl-XL dual ligand based on ILOE experiments and using suitable2A4 and BH3-I’ analogs.

With the above considerations, BH3-I 1 compound could use its carboxylic moiety to anchor the aminogroup of a linker, while a carboxylic containing 2A4 analog could also be reached through a peptide linkageas depicted in Figure 29. In this design all reagents are commercially available and a simple linkerexploration can be achieved with limited chemical hassle.


1.3 BIBLIOGRAPHY

1. Danial, N.N. and S.J. Korsmeyer, Cell death: Critical control points. Cell, 2004. 116(2): p. 205-219.2. Kaufmann, S.H. and M.O. Hengartner, Programmed cell death: alive and well in the new

millennium. Trends In Cell Biology, 2001. 11(12): p. 526-534.3. Brenner, C. and G. Kroemer, Apoptosis - Mitochondria - the death signal integrators. Science,

2000. 289(5482): p. 1150-1151.4. Deveraux, Q.L., et al., X-linked IAP is a direct inhibitor of cell-death proteases. Nature, 1997.

388(6639): p. 300-304.5. Huang, Y.H., et al., Structural basis of caspase inhibition by XIAP: Differential roles of the linker

versus the BIR domain. Cell, 2001. 104(5): p. 781-790.6. Riedl, S.J., et al., Structural basis for the inhibition of caspase-3 by XIAP. Cell, 2001. 104(5): p.

791-800.7. Chai, J.J., et al., Structural and biochemical basis of apoptotic activation by Smac/DIABLO. Nature,

2000. 406(6798): p. 855-862.8. LaCasse, E.C., et al., The inhibitors of apoptosis (IAPs) and their emerging role in cancer.

Oncogene, 1998. 17(25): p. 3247-3259.9. Fesik, S.W., Promoting apoptosis as a strategy for cancer drug discovery. Nature Reviews Cancer,

2005. 5(11): p. 876-885.10. Sun, C.H., et al., NMR structure and mutagenesis of the third Bir domain of the inhibitor of

apoptosis protein XIAP. Journal Of Biological Chemistry, 2000. 275(43): p. 33777-33781.11. D'Silva, L., et al., Monitoring the effects of antagonists on protein-protein interactions with NMR

spectroscopy. J Am Chem Soc, 2005. 127(38): p. 13220-6.12. Arac, D., T. Murphy, and J. Rizo, Facile detection of protein-protein interactions by one-

dimensional NMR spectroscopy. Biochemistry, 2003. 42(10): p. 2774-80.13. Bogan, A.A. and K.S. Thorn, Anatomy of hot spots in protein interfaces. J Mol Biol, 1998. 280(1): p.

1-9.14. Tsai, C.J., et al., Studies of protein-protein interfaces: A statistical analysis of the hydrophobic

effect. Protein Science, 1997. 6(1): p. 53-64.15. Liu, Z.H., et al., Structural basis for binding of Smac/DIABLO to the XIAP BIR3 domain. Nature,

2000. 408(6815): p. 1004-1008.16. Kegg, Kyoto Encyclopedia of Genes and Genomes. http://www.genome.jp/kegg/.17. Duewel, H., et al., Incorporation of trifluoromethionine into a phage lysozyme: implications and a

new marker for use in protein 19F NMR. Biochemistry, 1997. 36(11): p. 3404-16.18. Duewel, H.S., et al., Elucidation of solvent exposure, side-chain reactivity, and steric demands of

the trifluoromethionine residue in a recombinant protein. Biochemistry, 2001. 40(44): p. 13167-76.19. Eichler, J.F., et al., Biosynthetic incorporation of fluorohistidine into proteins in E. coli: a new probe

of macromolecular structure. Chembiochem, 2005. 6(12): p. 2170-3.20. Kim, H.W., et al., The specific incorporation of labelled aromatic amino acids into proteins through

growth of bacteria in the presence of glyphosate. Application to fluorotryptophan labelling to theH(+)-ATPase of Escherichia coli and NMR studies. FEBS Lett, 1990. 272(1-2): p. 34-6.

21. Gardner, K.H., M.K. Rosen, and L.E. Kay, Global folds of highly deuterated, methyl-protonatedproteins by multidimensional NMR. Biochemistry, 1997. 36(6): p. 1389-401.

22. Goto, N.K., et al., A robust and cost-effective method for the production of Val, Leu, Ile (delta 1)methyl-protonated 15N-, 13C-, 2H-labeled proteins. J Biomol NMR, 1999. 13(4): p. 369-74.

23. Lee, A.L., J.L. Urbauer, and A.J. Wand, Improved labeling strategy for 13C relaxationmeasurements of methyl groups in proteins. J Biomol NMR, 1997. 9(4): p. 437-40.

24. Rajesh, S., et al., A novel method for the biosynthesis of deuterated proteins with selectiveprotonation at the aromatic rings of Phe, Tyr and Trp. J Biomol NMR, 2003. 27(1): p. 81-6.


25. Kigawa, T., Y. Muto, and S. Yokoyama, Cell-free synthesis and amino acid-selective stable isotopelabeling of proteins for NMR analysis. J Biomol NMR, 1995. 6(2): p. 129-34.

26. Spirin, A.S., et al., A continuous cell-free translation system capable of producing polypeptides inhigh yield. Science, 1988. 242(4882): p. 1162-4.

27. Yabuki, T., et al., Dual amino acid-selective and site-directed stable-isotope labeling of the humanc-Ha-Ras protein by cell-free synthesis. J Biomol NMR, 1998. 11(3): p. 295-306.

28. Chong, S., et al., Utilizing the C-terminal cleavage activity of a protein splicing element to purifyrecombinant proteins in a single chromatographic step. Nucleic Acids Res, 1998. 26(22): p. 5109-15.

29. Dawson, P.E. and S.B. Kent, Synthesis of native proteins by chemical ligation. Annu Rev Biochem,2000. 69: p. 923-60.

30. Xu, R., et al., Chemical ligation of folded recombinant proteins: segmental isotopic labeling ofdomains for NMR studies. Proc Natl Acad Sci U S A, 1999. 96(2): p. 388-93.

31. Chin, J.W., et al., An expanded eukaryotic genetic code. Science, 2003. 301(5635): p. 964-967.32. Chin, J.W., et al., Progress toward an expanded eukaryotic genetic code. Chemistry & Biology,

2003. 10(6): p. 511-519.33. Pan, P., E. Woehl, and M.F. Dunn, Protein architecture, dynamics and allostery in tryptophan

synthase channeling. Trends Biochem Sci, 1997. 22(1): p. 22-7.34. Kanehisa, M. and S. Goto, KEGG: kyoto encyclopedia of genes and genomes. Nucleic Acids Res,

2000. 28(1): p. 27-30.35. Scott, F.L., et al., XIAP inhibits caspase-3 and -7 using two binding sites: evolutionarily conserved

mechanism of IAPs. Embo J, 2005. 24(3): p. 645-55.36. Shin, H., et al., The BIR domain of IAP-like protein 2 is conformationally unstable: implications for

caspase inhibition. Biochem J, 2005. 385(Pt 1): p. 1-10.37. Clinton, B., Successful presidential campaign. 1992.38. Woodward, C., N. Carulla, and G. Barany, Native state hydrogen-exchange analysis of protein

folding and protein motional domains, in Energetics Of Biological Macromolecules, Pt E. 2004. p.379-400.

39. Kessler, H., Conformation and Biological Activity of Cyclic Peptides. Angewandte ChemieInternational Edition in English, 1982. 21(7): p. 512-523.

40. Pervushin, K., et al., Attenuated T2 relaxation by mutual cancellation of dipole-dipole coupling andchemical shift anisotropy indicates an avenue to NMR structures of very large biologicalmacromolecules in solution. Proc Natl Acad Sci U S A, 1997. 94(23): p. 12366-71.

41. Ollerenshaw, J.E., V. Tugarinov, and L.E. Kay, Methyl TROSY: explanation and experimentalverification. Magnetic Resonance In Chemistry, 2003. 41(10): p. 843-852.

42. Pervushin, K., et al., Transverse relaxation-optimized spectroscopy (TROSY) for NMR studies ofaromatic spin systems in C-13-labeled proteins. Journal Of The American Chemical Society, 1998.120(25): p. 6394-6400.

43. Palmer, A.G., 3rd, NMR characterization of the dynamics of biomacromolecules. Chem Rev, 2004.104(8): p. 3623-40.

44. Bann, J.G., et al., Real-time and equilibrium (19)F-NMR studies reveal the role of domain-domaininteractions in the folding of the chaperone PapD. Proc Natl Acad Sci U S A, 2002. 99(2): p. 709-14.

45. Dalvit, C., et al., A general NMR method for rapid, efficient, and reliable biochemical screening. JAm Chem Soc, 2003. 125(47): p. 14620-5.

46. Bohm, H.J., et al., Fluorine in medicinal chemistry. Chembiochem, 2004. 5(5): p. 637-43.47. Khodursky, A.B., et al., DNA microarray analysis of gene expression in response to physiological

and genetic changes that affect tryptophan metabolism in Escherichia coli. Proc Natl Acad Sci U SA, 2000. 97(22): p. 12170-5.


48. Borden, K.L., P. Beckmann, and A.N. Lane, Determination of the orientations of tryptophananalogues bound to the trp repressor and the relationship to activation. Eur J Biochem, 1991.202(2): p. 459-70.

49. Ramesh, V., et al., NMR studies of the mode of binding of corepressors and inducers toEscherichia coli trp repressor. Eur J Biochem, 1996. 235(3): p. 804-13.

50. Marmorstein, R.Q. and P.B. Sigler, Stereochemical effects of L-tryptophan and its analogues on trprepressor's affinity for operator-DNA. J Biol Chem, 1989. 264(16): p. 9149-54.

51. Matchett, W.H., Inhibition of tryptophan synthetase by indoleacrylic acid. J Bacteriol, 1972. 110(1):p. 146-54.

52. Browne, D.T., G.L. Kenyon, and G.D. Hegeman, Incorporation Of Monofluorotryptophans IntoProtein During Growth Of Escherichia-Coli. Biochemical And Biophysical ResearchCommunications, 1970. 39(1): p. 13-&.

53. Gerig, J.T., Fluorine NMR of Proteins. Prog Nuc Magn Spectr, 1994. 26: p. 293-370.54. Sattler, M., et al., Structure of Bcl-x(L)-Bak peptide complex: Recognition between regulators of

apoptosis. Science, 1997. 275(5302): p. 983-986.55. Lugovskoy, A.A., et al., A novel approach for characterizing protein ligand complexes: molecular

basis for specificity of small-molecule Bcl-2 inhibitors. J Am Chem Soc, 2002. 124(7): p. 1234-40.56. Degterev, A., et al., Identification of small-molecule inhibitors of interaction between the BH3

domain and Bcl-xL. Nat Cell Biol, 2001. 3(2): p. 173-82.57. Medek, A., et al., The use of differential chemical shifts for determining the binding site location and

orientation of protein-bound ligands. Journal Of The American Chemical Society, 2000. 122(6): p.1241-1242.

58. Yin, H., et al., Terephthalamide derivatives as mimetics of helical peptides: disruption of the Bcl-x(L)/Bak interaction. J Am Chem Soc, 2005. 127(15): p. 5463-8.

59. Rutledge, S.E., J.W. Chin, and A. Schepartz, A view to a kill: ligands for Bcl-2 family proteins. CurrOpin Chem Biol, 2002. 6(4): p. 479-85.

60. Wang, J.L., et al., Structure-based discovery of an organic compound that binds Bcl-2 protein andinduces apoptosis of tumor cells. Proc Natl Acad Sci U S A, 2000. 97(13): p. 7124-9.

61. Enyedy, I.J., et al., Discovery of small-molecule inhibitors of Bcl-2 through structure-basedcomputer screening. J Med Chem, 2001. 44(25): p. 4313-24.

62. Kutzki, O., et al., Development of a potent Bcl-x(L) antagonist based on alpha-helix mimicry. J AmChem Soc, 2002. 124(40): p. 11838-9.

63. Yin, H., et al., Terphenyl-Based Bak BH3 alpha-helical proteomimetics as low-molecular-weightantagonists of Bcl-xL. J Am Chem Soc, 2005. 127(29): p. 10191-6.

64. Ernst, J.T., et al., Design and application of an alpha-helix-mimetic scaffold based on anoligoamide-foldamer strategy: antagonism of the Bak BH3/Bcl-xL complex. Angew Chem Int EdEngl, 2003. 42(5): p. 535-9.

65. Chin, J.W. and A. Schepartz, Design and Evolution of a Miniature Bcl-2 Binding Protein. AngewChem Int Ed Engl, 2001. 40(20): p. 3806-3809.

66. Hajduk, P.J., J.R. Huth, and S.W. Fesik, Druggability indices for protein targets derived from NMR-based screening data. J Med Chem, 2005. 48(7): p. 2518-25.

67. Muchmore, S.W., et al., X-ray and NMR structure of human Bcl-x(L), an inhibitor of programmedcell death. Nature, 1996. 381(6580): p. 335-341.

68. Pellecchia, M., D.S. Sem, and K. Wuthrich, NMR in drug discovery. Nat Rev Drug Discov, 2002.1(3): p. 211-9.

69. Meyer, B. and T. Peters, NMR Spectroscopy techniques for screening and identifying ligandbinding to protein receptors. Angewandte Chemie-International Edition, 2003. 42(8): p. 864-890.

70. London, R.E., Theoretical analysis of the inter-ligand overhauser effect: a new approach formapping structural relationships of macromolecular ligands. J Magn Reson, 1999. 141(2): p. 301-11.


71. Becattini, B. and M. Pellecchia, SAR by ILOEs: an NMR-based approach to reverse chemicalgenetics. Chemistry, 2006. 12(10): p. 2658-62.

72. Fejzo, J., et al., The SHAPES strategy: an NMR-based approach for lead generation in drugdiscovery. Chem Biol, 1999. 6(10): p. 755-69.

73. Hajduk, P.J., et al., Privileged molecules for protein binding identified from NMR-based screening.J Med Chem, 2000. 43(18): p. 3443-7.

74. Becattini, B. and M. Pellecchia, SAR by ILOEs: An NMR-Based Approach to Reverse ChemicalGenetics. Chemistry, 2005.

nmr in drug discovery. from screening to structure-based design of antitumoral...

Documents